Abstract: The Cold-start problem considered as one of the most common limitations of Recommender Systems (RSs) which indicates to situation when a new user or items lastly combined to the system. We focus in this study on set of the users who did not receive attention by researchers about their preferences in future, so they often have been ignored. Highlight on their interests may improve the recommendation systems; we utilize the social relation information and social network analysis to alleviating the new users problem by exploiting the preferences of influential nodes to recommend items for new users. So, the main contribution in this research is employing the influential nodes to alleviate the cold start problem. Top-M influential nodes have been assigned from the social communities using closeness and degree centrality measure and then suggest the most popular interest items of those nodes to new users. The performance of our approach has been applied using two data sets: Lastfm published by Group lens research and CiaoDVD published by LibRec research group. The experimental results show that exploiting influential nodes in cold start issue improve the recommendation system.
Mohsin Hasan Hussein, Huda Naji Nawaf and Wesam Bhaya, 2016. Influential Nodes Based Alleviation of User Cold-Start Problem in Recommendation System. Research Journal of Applied Sciences, 11: 1107-1114.